Worked on the VCF straight 8 that was acting up again. It would load the first instruction to execute when you hit the start key but didn't load the next instruction after updating the program counter. Kept reexecuting the first instruction. TLDR: Found some signals that weren't quite right but fixing them didn't help. Found actual bad diode that was causing the failure but then diagnostics found another fault. Whike working on that memory stopped working when machine on for a while. Ran out of time to determine why so still not usable. Started browsing signals. PDP-8 signals are referred to as 0V and -3V. They actually are one diode drop below -3V. The F(1) signals which indicates the machine is in the fetch state was only going to about -1.8V. Normally this is caused by a leaky diode on logic being driven by the signal. Went through the schematics trying to find everywhere the signal went. Pulled 12 boards and tested diodes. Found and replaced one with high forward voltage but none measured low forward voltage indicating they are leaky. Reverse test would be better but components on the boards make some diodes read as leaky and hard to be sure you have gotten good connection through the surface oxides. Good diode and poor connection read the same on the meter. F(1) drives the light on the front panel through a drive transistor. Found that if I put a piece of tape on the edge connector to open that connection the signal level returned to proper. The manual doesn't have schematic for the front panel so need to figure that out. Much time spent and it didn't fix the problem. Will need to make measurements on my machine so I can compare with VCF and determine real cause of signal level issue. Suspect its not really the front panel but removing that load allowed signal to go to proper level. Looking at more signals found PA30 J output F set which sets the fetch flip flop also wasn't going as low as it should. I put a replacement S111 in and that fixed it. No change to symptoms. Later after I fixed the real problem I was going to swap the card back and see if it caused problems. I couldn't find a S111 card around. Wondered if my notes were a hallucination. Did find a R111 around but thought I had pulled a S111. The difference between the cards is the pull down resistors to -15V are changed from 7.5K on R111 to 3K on S111. This improves the fall time on the signal and gives more low drive (4mA vs 1.6 mA). The increased drive may be hiding a leaky diode. Don't have specifications for D-664 diode to know expected leakage. Typical small signals diodes are less than a microamp. I checked old pictures of the machine and PA30 had a R111. The early manual had R111 shown in the schematic but the card layout chart showed S111. The later manual had S111 in both locations. Next found spec cycle signal was at -1.4V. Checked cards signal goes to and found PA32 had a diode where the forward voltage measured .1V. Replaced diode and signal now had proper level and it started fetching instructions properly. Ran maindec-801-1-pm and it halted at 1062. This test was checking the proper return address got stored on an interrupt. Found the previous tests return address was at address 0. Single cycle stepping showed that it wrote the return address to location 4 instead of 0. The memory address register bit 9 didn't clear when it was supposed to. Pulled the R211 card and checked the diodes and found one with 1V forward drop and replaced it. Didn't fix it. At this point memory stopped working. All reads returned 0. Found if I turned the machine off for a while memory would operates for a while. Test still failed. Swapping in a spare R211 fixed the problem. The memory signals seemed to be operating normally when reads weren't working. The sense amplifier E pin didn't show a pulse. All memory locations and all bits don't work. All power supply voltages seemed reasonable and didn't change significantly between working and not working. Wasn't able to determine what the problem was before time to pack up.